Volume 26, Issue 3, Pages 523-538.e5 (September 2017) Distinct Circadian Signatures in Liver and Gut Clocks Revealed by Ketogenic Diet  Paola Tognini, Mari Murakami, Yu Liu, Kristin L. Eckel-Mahan, John C. Newman, Eric Verdin, Pierre Baldi, Paolo Sassone-Corsi  Cell Metabolism  Volume 26, Issue 3, Pages 523-538.e5 (September 2017) DOI: 10.1016/j.cmet.2017.08.015 Copyright © 2017 Elsevier Inc. Terms and Conditions

Cell Metabolism 2017 26, 523-538.e5DOI: (10.1016/j.cmet.2017.08.015) Copyright © 2017 Elsevier Inc. Terms and Conditions

Figure 1 Heatmaps, KEGG Pathway, and Phase Lag Analyses of the Diurnal Transcriptome in Liver and Gut upon KD (A) Venn diagram representing the number of genes rhythmic only in KD, only in CC, and in both conditions in the liver (n = 3 per time point, per group, p < 0.01, false discovery rate [FDR] < 0.1). (B) Heatmaps representing the genes diurnal in both conditions, KD only, and CC only in liver (n = 3 per time point, per group, p < 0.01, FDR < 0.1). (C) KEGG pathway analysis in liver. The numbers in the pie charts and tables represent the number of genes enriched in the specific KEGG pathway. (D) Radar plots representing the phase lag of genes exclusively diurnal in KD or CC in hepatic tissue (Anderson-Darling test; phase distributions significance, p < 0.0001). (E) Venn diagram representing the number of genes oscillating only in KD, only in CC, and in both conditions in the intestine (n = 3 per time point, per group, p < 0.01, FDR < 0.1). (F) Heatmaps representing the genes diurnal in both conditions, KD only, and CC only in the gut (n = 3 per time point, per group, p < 0.01, FDR < 0.1). (G) KEGG pathway analysis in the intestine. The numbers in the pie charts and tables represent the number of genes enriched in the specific KEGG pathway. (H) Radar plots representing the phase lag of genes exclusively diurnal in KD or CC in ileal cells (Anderson-Darling test; phase distribution significance in IECs, KD versus CC, p < 0.0001; comparisons not shown; phase distribution difference upon KD, liver versus IECs, p < 0.0001; phase distribution difference upon CC, liver versus IECs, p < 0.0001). Cell Metabolism 2017 26, 523-538.e5DOI: (10.1016/j.cmet.2017.08.015) Copyright © 2017 Elsevier Inc. Terms and Conditions

Figure 2 Metabolite Levels and Circadian Transcriptional Signature in Liver versus IECs upon KD (A) Hepatic total cholesterol levels at ZT0 and ZT12 in CC- and KD-fed mice (n = 4 per time point, per group; two-way ANOVA, Holm-Sidak post hoc; comparisons for factor, diet within ZT, ∗p < 0.05; ZT within KD, #p < 0.05). (B) Free fatty acid levels in the liver of CC- and KD-fed mice at ZT8 and ZT20 (n = 4 per time point, per group; two-way ANOVA, Holm-Sidak post hoc; comparisons for factor, diet within ZT, ∗p < 0.05). (C) Intestinal total cholesterol levels at ZT0 and ZT12 in CC- and KD-fed mice (n = 4 per time point, per group; two-way ANOVA, no significant difference). (D) Free fatty acid levels in the ileum of CC- and KD-fed mice at ZT8 and ZT20 (n = 4 per time point, per group; two-way ANOVA, Holm-Sidak post hoc; comparisons for factor, diet within ZT, ∗p < 0.05; ZT within KD, #p < 0.05). Error bars represent SEM. (E) Venn diagram representing the transcripts oscillating in “liver only,” “both tissues,” and “IECs only” upon KD feeding (n = 3 per time point, per group, p < 0.01, FDR < 0.1). (F) Heatmaps representing the genes diurnal in both tissues, liver only, and intestine only upon KD (n = 3 per time point, per group, p < 0.01, FDR < 0.1). (G) Analysis of the rhythmic transcription factors (p < 0.01) in animals fed KD, exclusively in the liver, in both tissues, or exclusively in the gut. (H) Transcription factor binding site (TFBS) analysis in the liver of KD-fed mice. In the graph is reported the –log (pValue, 10). The first four sites are enriched exclusively in the liver upon KD. CLOCK:BMAL1 binding site is not exclusive, although it is highly enriched in KD liver, as shown by the value of the –log (pValue). (I) Transcription factor binding site (TFBS) analysis in the IECs of KD-fed mice. In the graph is reported the –log (pValue, 10). Cell Metabolism 2017 26, 523-538.e5DOI: (10.1016/j.cmet.2017.08.015) Copyright © 2017 Elsevier Inc. Terms and Conditions

Figure 3 Clock-Controlled Gene Expression and BMAL1 Chromatin Recruitment on Specific Target Promoters (A) Expression of clock output genes Dbp, Nampt, and Pnpla2 measured by qPCR in the liver (n = 5 per time point, per group; two-way ANOVA, Holm-Sidak post hoc; comparisons for factor, diet within ZT, ∗p < 0.05). (B) qPCR- ChIP results showing BMAL1 chromatin recruitment on the E-boxes of the Dbp, Nampt, and Pnpla2 gene promoters in the liver (n = 5 per time point, per group; two-way ANOVA, Holm-Sidak post hoc; comparisons for factor, diet within ZT, ∗p < 0.05). (C) Expression of clock output genes Dbp, Nampt, and Pnpla2 measured by qPCR in IECs (n = 5 per time point, per group; two-way ANOVA, Holm-Sidak post hoc; comparisons for factor, diet within ZT, ∗p < 0.05). (D) ChIP results showing BMAL1 chromatin recruitment on the E-boxes of Dbp, Nampt, and Pnpla2 gene promoters in IECs (n = 5 per time point, per group; two-way ANOVA, Holm-Sidak post hoc; comparisons for factor, diet within ZT, no significant difference). IgG represents ChIP experiment performed with an isotype-matched control immunoglobulin (normal rabbit IgG) to BMAL1. Error bars represent SEM. Cell Metabolism 2017 26, 523-538.e5DOI: (10.1016/j.cmet.2017.08.015) Copyright © 2017 Elsevier Inc. Terms and Conditions

Figure 4 Tissue-Specific Circadian PPARα Signature in Liver and Intestine upon KD (A) Western blot analysis of PPARα in liver nuclear extract. On the left, representative blot; on the right, graph reporting the average of the blot band density (n = 3 per time point, per group; two-way ANOVA, Holm-Sidak post hoc; comparisons for factor, diet within ZT, ∗p < 0.05). (B) Expression of the PPARα target genes Hmgcs2, Acot2, Cpt1a, and Angptl4 in liver extract (n = 5 per time point, per group; two-way ANOVA, Holm-Sidak post hoc; comparisons for factor, diet within ZT, ∗p < 0.05). (C) Liver PPARα target gene expression in mice fed CC or KD for 4 weeks and treated with the specific PPARα inhibitor GW6471 (n = 3–4 per time point, per group; two-way ANOVA, Holm-Sidak post hoc; comparisons for factor, diet within ZT, ∗p < 0.05). (D) Nuclear PPARα protein accumulation in IECs. On the left, representative blot; on the right, graph reporting the average of the blot band density (n = 3 per time point, per group; two-way ANOVA, Holm-Sidak post hoc; comparisons for factor, diet within ZT, ∗p < 0.05). (E) PPARα target gene expression of Hmgcs2, Acot2, Cpt1a, and Angptl4 in IECs (n = 5 per time point, per group; two-way ANOVA, Holm-Sidak post hoc; comparisons for factor, diet within ZT, ∗p < 0.05). (F) Ileum PPARα target gene expression in mice fed CC or KD for 4 weeks and treated with the specific PPARα inhibitor GW6471 (n = 3–4 per time point, per group; two-way ANOVA, Holm-Sidak post hoc; comparisons for factor, diet within ZT, ∗p < 0.05). Error bars represent SEM. Cell Metabolism 2017 26, 523-538.e5DOI: (10.1016/j.cmet.2017.08.015) Copyright © 2017 Elsevier Inc. Terms and Conditions

Figure 5 KD-Driven Oscillation in Serum βOHB and Daily Changes in Epigenetic Marks (A) Liver ChIP of acetylated (Acetyl) lysine (K) 9-14 on histone H3. qPCR showing the relative abundance of H3 Acetyl(K9-K14) on the PPAR-responsive element (PPRE) of specific promoter: Hmgcs2, Acot2, Cpt1a, and Angptl4 (n = 4–6 per time point, per group; two-way ANOVA, Holm-Sidak post hoc; comparisons for factor, diet within ZT, ∗p < 0.05). (B) H3 Acetyl(K9-K14) abundance on the PPRE of Hmgcs2, Acot2, Cpt1a, and Angptl4 promoter in the intestine (n = 4 per time point, per group; two-way ANOVA, Holm-Sidak post hoc; comparisons for factor, diet within ZT, ∗p < 0.05; ZT within KD, #p < 0.05). (C) Serum βOHB levels along the diurnal cycle in CC- and KD-fed mice (n = 5 per time point, per group; two-way ANOVA, Holm-Sidak post hoc; comparisons for factor, diet within ZT, ∗p < 0.05). (D) Hepatic βOHB levels along the diurnal cycle in CC- and KD-fed mice (n = 3 per time point, per group; two-way ANOVA, Holm-Sidak post hoc; comparisons for factor, diet within ZT, ∗p < 0.05). (E) Intestinal βOHB levels along the diurnal cycle in CC- and KD-fed mice (n = 3 per time point, per group; two-way ANOVA, Holm-Sidak post hoc; comparisons for factor, diet within ZT, ∗p < 0.05). (F) HDAC activity from liver nuclear extract at ZT8 and ZT20 (n = 5 per time point, per group; two-way ANOVA, Holm-Sidak post hoc; comparisons for factor, diet within ZT, no significant difference; ZT within KD, #p < 0.05). (G) HDAC activity in intestinal nuclear extract at ZT8 and ZT20 (n = 3–5 per time point, per group; two-way ANOVA, Holm-Sidak post hoc; comparisons for factor, diet within ZT, ∗p < 0.05; ZT within KD, #p < 0.05). Error bars represent SEM. Cell Metabolism 2017 26, 523-538.e5DOI: (10.1016/j.cmet.2017.08.015) Copyright © 2017 Elsevier Inc. Terms and Conditions

Figure 6 Liver and Intestine Response to Perturbation in Diet Composition (A) qPCR of genes involved in the ketogenic response, Hmgcs2, Acot2, and Angptl4, in the liver of mice fed 4-week KD or CC and administered with fructose oral gavage (4 g/Kg) once a day, per 7 days during the last week of dietary regimen (n = 5–6 per time point, per group; two-way ANOVA, Holm-Sidak post hoc; comparisons for factor, diet within ZT, ∗p < 0.05). (B) Hmgcs2, Acot2, and Angptl4 gene expression in the liver of animals fed 4-week KD or CC and simultaneously treated with 30% sucrose in the drinking water (n = 5 per time point, per group; two-way ANOVA, Holm-Sidak post hoc; comparisons for factor, diet within ZT, ∗p < 0.05). (C) Intestinal gene expression of Hmgcs2, Acot2, and Angptl4 in mice subjected to fructose oral gavage (n = 5–6 per time point, per group; two-way ANOVA, Holm-Sidak post hoc; comparisons for factor, diet within ZT, ∗p < 0.05). (D) qPCR analysis of Hmgcs2, Acot2, and Angptl4 genes in the intestine of animals treated with 30% sucrose in the drinking water (n = 5 per time point, per group; two-way ANOVA, Holm-Sidak post hoc; comparisons for factor, diet within ZT, ∗p < 0.05). Error bars represent SEM. Cell Metabolism 2017 26, 523-538.e5DOI: (10.1016/j.cmet.2017.08.015) Copyright © 2017 Elsevier Inc. Terms and Conditions

Figure 7 Schematic Representation of KD-Dependent Diurnal Signature in Liver and Intestine A very-low-carbohydrate KD is able to reprogram peripheral clocks in liver and intestine through the activation of distinct transcription factors in a diurnal manner. KD enhances BMAL1 chromatin recruitment in the liver, resulting in an increase in the amplitude of oscillation of CCGs. KD activates the PPARα pathway both in liver and intestine but is highly diurnal only in the gut. The activation of ketogenesis increases serum βOHB concentration that oscillates in a diurnal fashion. Through circulation, βOHB is distributed to all the tissues as an energy source. The effect of βOHB is primarily evident in the intestine, where the metabolite oscillates with the same phase of serum βOHB, and it leads to regulated HDAC activity in a time-of-the-day-dependent fashion. This results in the increase of histone H3 acetylation on metabolic gene promoters at specific ZTs, contributing to rhythmic gene expression. Indeed, PPARα targets display a unique diurnal profile that nicely parallels PPARα nuclear accumulation and serum/intestinal βOHB daily changes. Cell Metabolism 2017 26, 523-538.e5DOI: (10.1016/j.cmet.2017.08.015) Copyright © 2017 Elsevier Inc. Terms and Conditions